Skip to navigation

Threat reportVulnerabilityTL-2026-1941

N-able N-central Authentication Bypass Flaws (CVE-2026-18556, CVE-2026-18577) Actively Exploited for Admin Access and Cloudflare Tunnel Persistence

highACTIVE

N-able released a second hotfix (build 2026.3.1.10) for its N-central RMM platform after threat actors actively exploited two related authentication-bypass/account-takeover vulnerabilities (CWE-288) to gain unauthenticated administrative access and establish Cloudflare Tunnel persistence on managed endpoints. CVE-2026-18577 is an incomplete fix for the earlier CVE-2026-18556; both are CISA KEV-listed as actively exploited, with Huntress observing one compromised self-hosted partner account reach nine downstream organizations.

Key facts for N-central Cloudflare Tunnel persistence

IdentifiersTL-2026-1941, CVE-2026-18556, CVE-2026-18577
SeverityHIGH
CVSS8.2 (CVSS:4.0/AV:N/AC:H/AT:N/PR:N/UI:N/VC:H/VI:N/VA:N/SC:L/SI:L/SA:L/E:A)
Exploitation statusExploited in the wild; Listed in CISA KEV: CVE-2026-18577 (added )
EPSSCVE-2026-18577: 1.4% (71.3th percentile); CVE-2026-18556: 0.2% (19.1th percentile)
Affected productsN-able N-central
Weaknesses (CWE)CWE-288
MalwarePhantomKiller
ToolsAnyDesk, CloudFlare Tunnel, HopToDesk, RustDesk, SimpleHelp, Tactical RMM, TeamViewer
MITRE ATT&CK techniquesT1018 Remote System Discovery, T1036.005 Match Legitimate Resource Name or Location, T1057 Process Discovery, T1059.001 PowerShell, T1069.002 Domain Groups, T1087.002 Domain Account, T1090 Proxy, T1098 Account Manipulation, T1136.002 Domain Account, T1190 Exploit Public-Facing Application, T1199 Trusted Relationship, T1211 Exploitation for Stealth and 6 more in the attack flow
MotivationUNKNOWN
Target sectorsinformation technology, managed service providers
Target regionsGlobal
Detection rules9 (1 shown free)
Sources22 cited sources

Detection coverage for N-central Cloudflare Tunnel persistence

As of 2026-08-08, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-1941 across Splunk SPL, Microsoft KQL and Sigma, covering 35 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.

Free detection: Rogue Cloudflare Tunnel Service Masquerading as MicrosoftEdgeUpdate64.exe / msmp.exe

Detects registration of a new Windows service running cloudflared (Cloudflare Tunnel) under a masqueraded binary name (MicrosoftEdgeUpdate64.exe, msmp.exe) outside the legitimate Edge installation path — the confirmed persistence mechanism that survived revocation of the attackers' N-central access, giving them a firewall-evading outbound C2 channel.

  • Languages: Splunk SPL, Microsoft KQL, Sigma
  • ATT&CK: T1543.003, T1572, T1036.005
  • Data sources: windows_security, sysmon, service_control_manager
  • Severity: critical
  • Confidence: high

Splunk SPL

index=wineventlog sourcetype="WinEventLog:System" EventCode=7045
(ImagePath="*cloudflared*" OR ImagePath="*MicrosoftEdgeUpdate64.exe*" OR ImagePath="*msmp.exe*")
NOT ImagePath="*\\Program Files (x86)\\Microsoft\\Edge*"
| table _time Computer_Name Service_Name ImagePath Account_Name
| sort -_time

Microsoft KQL

DeviceProcessEvents
| where Timestamp > ago(24h)
| where (ProcessCommandLine has "cloudflared" or FileName in~ ("MicrosoftEdgeUpdate64.exe","msmp.exe"))
| where FolderPath !startswith @"C:\Program Files (x86)\Microsoft\Edge"
| project Timestamp, DeviceName, FileName, FolderPath, ProcessCommandLine, InitiatingProcessAccountName

Sigma

title: Cloudflare Tunnel Service Registered Under Masqueraded Microsoft Edge Update Binary Name
id: 7a1e2f3c-4b5d-4e6f-8a9b-1c2d3e4f5a07
status: experimental
description: Detects a new Windows service whose ImagePath references cloudflared or the masqueraded names MicrosoftEdgeUpdate64.exe/msmp.exe outside the legitimate Microsoft Edge install directory — the confirmed cloudflared persistence mechanism observed post N-central compromise.
references:
  - https://thehackernews.com/2026/08/n-central-attackers-reach-managed.html
  - https://www.sophos.com/en-us/blog/nable-ncentral-exploitation-results-in-rmm-tool-deployment
author: AII-Detector — ThreadLinqs Intelligence
date: 2026-08-08
tags:
  - attack.persistence
  - attack.command-and-control
  - attack.t1543.003
  - attack.t1572
  - attack.t1036.005
logsource:
  category: service_creation
  product: windows
detection:
  selection:
    EventID: 7045
  filter_path:
    ImagePath|contains: 'cloudflared'
  filter_masquerade:
    ImagePath|contains:
      - 'MicrosoftEdgeUpdate64.exe'
      - 'msmp.exe'
  filter_legit:
    ImagePath|contains: '\Program Files (x86)\Microsoft\Edge'
  condition: selection and (filter_path or filter_masquerade) and not filter_legit
falsepositives:
  - Legitimate enterprise use of Cloudflare Tunnel for approved remote-access/zero-trust deployments (allow-list known-good service names/paths)
level: critical

Known false positives:

  • Legitimate, sanctioned Cloudflare Tunnel deployments for zero-trust remote access — allow-list known-good service names and binary paths/hashes

Tuning notes: Allow-list any organizationally-approved cloudflared service by binary hash/path; treat any hit on the MicrosoftEdgeUpdate64.exe/msmp.exe masquerade filter outside the real Edge directory as high-confidence regardless of the cloudflared string match.

8 more detection rules (Splunk SPL, Microsoft KQL, Sigma) · unlock with TL Creds. Sign up free to unlock the detection rules or query them with the Threadlinqs MCP server (Purple)
  • N-central Trusted-Relationship Abuse — Single Partner Account Pivoting to Multiple Downstream Organizations · unlock with TL Creds
  • Domain Reconnaissance Command Sequence Following N-central Compromise (net user / net group / nltest) · unlock with TL Creds
  • Redundant Remote-Access Tool Sprawl + Rogue TacticalRMM Callback to [indicator] · unlock with TL Creds
  • N-central Alternate-Path Auth Bypass Request Pattern (CVE-2026-18577 Incomplete-Patch Exploitation) · unlock with TL Creds
  • PhantomKiller BYOVD EDR-Evasion Tool — Kernel Driver Load and Security-Software Fingerprinting · unlock with TL Creds
  • Network Connection to Disclosed C2 IP/Domain Watchlist (Cloudflare Tunnel / QuickConnect Infrastructure) · unlock with TL Creds
  • N-central Auth Bypass Exploitation — Admin Access From Disclosed Attacker Infrastructure (CVE-2026-18556/CVE-2026-18577) · unlock with TL Creds
  • Rogue Domain Account Creation ("veeam") and Administrator Password Reset via Compromised N-central Access · unlock with TL Creds

Indicators of compromise for N-central Cloudflare Tunnel persistence

As of 2026-08-08, Threadlinqs Intelligence tracks 35 indicators of compromise for TL-2026-1941.

Sample indicators (3 of 35)

  • IP address: 173.249.252[.]176
  • Domain: mousears[.]synology[.]me
  • IP address: 173.249.252[.]200

Defanged, TLP:CLEAR. Full set of 35 indicators: unlock with TL Creds.

35 indicators of compromise · unlock with TL Creds. Compare plans

How N-central Cloudflare Tunnel persistence works

N-able's N-central remote monitoring and management (RMM) platform, used by managed service providers (MSPs) and enterprise IT teams to administer downstream customer environments, contains two related authentication-bypass vulnerabilities (CWE-288, Authentication Bypass Using an Alternate Path or Channel) enabling unauthenticated administrative account takeover. CVE-2026-18556 (CVSS 3.1 7.4 -- AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:N -- / CVSS 4.0 8.2) affected all N-central versions through 2026.1 and was addressed by an initial vendor patch in build 2026.2. That patch proved incomplete: CVE-2026-18577 (CVSS 2.0 7.6 / CVSS 3.1 8.1 -- AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H -- / CVSS 4.0 8.2) is a residual flaw affecting all N-central versions prior to build 2026.3.1.7, letting attackers route around the earlier fix via an alternate exploitation path and again obtain full, unauthenticated administrative access remotely -- both on cloud-hosted (NCOD) and on-premises deployments, described by researchers as effectively "trivial to exploit."

N-able first detected the activity on July 31, 2026 as a spike in anomalous licensing errors reported by on-premises customers, traced via its Adlumin MDR service to active exploitation of what would be designated CVE-2026-18577. On August 2, 2026 the vendor shipped Hotfix 1 (build 2026.3.1.7) and published its first advisory; Finland's national cyber security centre also issued a notice the same day. Huntress published a rapid-response writeup on August 3 confirming exploitation across multiple organizations -- in at least one case a single compromised self-hosted MSP partner account was used to reach nine downstream client organizations, one endpoint compromised per organization -- and found that as of 12:45 AM ET on August 3, 55.6% of internet-reachable cloud (NCOD) N-central servers were still unpatched; a subsequent same-day Huntress update reported patch compliance improving to 13.6% of internet-reachable cloud servers and 28.6% of self-hosted instances still unpatched/vulnerable. CISA added CVE-2026-18577 to its Known Exploited Vulnerabilities (KEV) catalog on August 3, 2026 (remediation due August 6) and CVE-2026-18556 on August 4, 2026 (remediation due August 7), directing federal civilian agencies to remediate under Binding Operational Directive 22-01. Sophos independently identified a compromised customer in its telemetry at approximately 08:00 UTC on August 3, and a second compromised organization on August 5. Continued monitoring of threat-actor behavior showed Hotfix 1's hardening was insufficient, and on August 6, 2026 N-able shipped Hotfix 2 (build 2026.3.1.10), stating explicitly it was required even for customers who had already applied Hotfix 1, alongside an expanded 10-IP indicator list and a downloadable IOC-detection service template for Windows endpoints.

Post-exploitation, attackers used unauthenticated administrative access to the N-central console to invoke the built-in Take Control remote-access feature and pivot into managed downstream endpoints, deliberately targeting high-value systems -- domain controllers, backup servers, and application servers -- rather than opportunistic access, effectively abusing the trusted MSP-to-client management channel itself as the lateral-movement path. Sophos telemetry documented attackers running "net user" and "net user /domain" account-enumeration commands, creating a new rogue domain account named "veeam," and resetting existing administrator passwords (account manipulation/persistence), then running reconnaissance commands "nltest /dclist" (domain-controller enumeration) and "net group \"domain admins\" /domain" (domain-admin group enumeration). Attackers piped "tasklist" output through "findstr ms" and "findstr soph" to specifically fingerprint running Microsoft Defender and Sophos agents, then used a Bring-Your-Own-Vulnerable-Driver-style EDR-evasion tool dubbed "PhantomKiller" (also observed renamed to 9.exe) -- which loaded a kernel driver (k.sys) staged inside C:\ProgramData\AnyDesk -- to terminate the Sophos File Scanner process (sophosfilescanner.exe) and impair endpoint defenses; Sophos tracks detection for this activity internally as CXmal/KillAV-BR. For persistence, actors registered a new Windows service to run a Cloudflare Tunnel (cloudflared), masquerading the binary under names such as "MicrosoftEdgeUpdate64.exe" and "msmp.exe" to blend in with legitimate Windows/Edge update processes; this gave them an outbound, firewall-evading channel that survived revocation of their access to the N-central server itself. Huntress and Sophos additionally observed a wide sprawl of legitimate commercial remote-access tools installed across compromised endpoints for redundant access -- AnyDesk, TeamViewer (via TeamViewer_Setup.exe / team.msi), RustDesk, SimpleHelp (service64off.exe), HopToDesk, and a rogue TacticalRMM v2.11.0 agent deployed via a PowerShell installer script (install_server.ps1) that dropped tacticalagent-v2.11.0-windows-amd64.exe, phoning home to attacker infrastructure at api.mendoratech.health. A suspicious svchost.exe binary was also found planted in a user's Documents folder, masquerading as a legitimate Windows process outside its expected System32 location. Disclosed indicators include ten IP addresses and three .synology.me/quickconnect.to domains, several of the IPs resolving to Mullvad and NordVPN commercial VPN exit nodes, consistent with deliberate origin obfuscation; one of the Mullvad-associated IPs had prior history of abuse for brute-forcing and spam. No public attribution to a specific threat actor or group has been made. N-able has stated a limited but unspecified number of customers were affected, and the incident is being cited by third-party analysts as an illustration of RMM/MSP platforms' outsized supply-chain blast radius -- a single authentication-bypass exploit against one MSP's management server can cascade into intrusions across every downstream client it manages.

MITRE ATT&CK techniques by tactic:

Discovery

T1018 Remote System Discovery; T1057 Process Discovery; T1069.002 Domain Groups; T1087.002 Domain Account; T1518.001 Security Software Discovery

Stealth

T1036.005 Match Legitimate Resource Name or Location; T1211 Exploitation for Stealth

Execution

T1059.001 PowerShell

Command and Control

T1090 Proxy; T1219 Remote Access Tools; T1572 Protocol Tunneling

Persistence

T1098 Account Manipulation; T1136.002 Domain Account; T1543.003 Windows Service

Initial Access

T1190 Exploit Public-Facing Application; T1199 Trusted Relationship

Lateral Movement

T1570 Lateral Tool Transfer

Defense Impairment

T1685 Disable or Modify Tools

Remediation for N-central Cloudflare Tunnel persistence

Patches

  • N-central 2026.3.1.7 (Hotfix 1, released 2026-08-02, mitigates CVE-2026-18577 but proved incomplete)
  • N-central 2026.3.1.10 (Hotfix 2, released 2026-08-06, additional hardening, required even if Hotfix 1 was already applied)

Immediate actions

  • Apply N-central Hotfix 2 (build 2026.3.1.10) on all on-premises deployments immediately -- Hotfix 1 (2026.3.1.7) alone is insufficient per N-able's August 6 advisory
  • Run N-able's custom IOC-detection service template (developer.n-able.com/n-central/recipes/cve-2026-18577-detection) against all N-central-managed Windows endpoints
  • Review N-central Take Control / remote-access session logs (ui_access_control.log) and Windows Event IDs 4102, 8192, 8193 for connections from the disclosed IP list, especially sessions targeting domain controllers, backup servers, or application servers outside normal support-ticket activity
  • Hunt for rogue Cloudflare Tunnel (cloudflared) Windows services and files masquerading as MicrosoftEdgeUpdate64.exe, msmp.exe, or svchost.exe outside System32 (e.g. user Documents folders)
  • Check C:\ProgramData\GetSupportService_N-Central\Logs\ for BASupSrvc_*.log.gz entries whose timestamps correlate with suspicious N-central sessions
  • Audit Active Directory for unexpected new domain accounts (e.g. accounts named to resemble backup/software vendors such as "veeam"), recent administrator password resets, and "net user"/"net group" domain-enumeration activity in command-line logging
  • Check for a suspicious kernel driver file named k.sys (or renamed loader 9.exe) under C:\ProgramData\AnyDesk and for processes/services matching the BYOVD-style EDR-evasion tool "PhantomKiller" (Sophos detection: CXmal/KillAV-BR)
  • Review default/support account activity (e.g. mspsupport@n-able.com / "MSP Support") for logins and Take Control sessions from unfamiliar source IPs or off-hours timing
  • Inventory all managed endpoints for unauthorized installs of AnyDesk, TeamViewer (TeamViewer_Setup.exe/team.msi), RustDesk, SimpleHelp (service64off.exe), HopToDesk, or TacticalRMM (particularly v2.11.0 / tacticalagent-v2.11.0-windows-amd64.exe deployed via install_server.ps1, phoning home to api.mendoratech.health)

Workarounds

  • Hosted/cloud (NCOD) N-central deployments receive fixes automatically; no customer action required
  • On-premises customers unable to patch immediately should isolate the N-central admin interface from untrusted networks, place it behind a reverse proxy/Cloudflare Access, and closely monitor Take Control session logs for anomalous activity
  • Consider taking self-hosted N-central instances temporarily offline in critical environments until patched

Longer-term hardening

  • Restrict N-central console/API access to trusted administrative source IPs via firewall allow-listing
  • Enforce phishing-resistant MFA/SSO on all N-central administrative and support accounts
  • Monitor outbound traffic for unauthorized tunneling tools (cloudflared) and unexpected Cloudflare Tunnel infrastructure connections
  • Establish continuous patch-compliance monitoring for RMM/MSP tooling given its downstream supply-chain blast radius across managed customer environments
  • Rotate N-central credentials, API keys, and integration secrets post-incident, and review Take Control session logs back to at least July 1, 2026
  • For MSPs: define a per-tenant client-notification SLA and provide clients evidence of patch status, IOC hunting results, and credential-rotation completion

Affected and fixed versions

  • N-able — N-central
    Vulnerable versions: through 2026.1 (CVE-2026-18556, patched in 2026.2 but incompletely); 2026.2 through 2026.3.1, prior to build 2026.3.1.7 (CVE-2026-18577)
    Fixed in: 2026.3.1.7 (Hotfix 1, partial mitigation); 2026.3.1.10 (Hotfix 2, full mitigation, required regardless of Hotfix 1 status)

Timeline of N-central Cloudflare Tunnel persistence

  • N-able detects a spike in anomalous licensing errors from on-premises customers, traced via its Adlumin MDR service to active exploitation of an N-central authentication-bypass flaw later designated CVE-2026-18577.
  • CVE-2026-18556 is publicly disclosed, describing the original N-central authentication-bypass-via-alternate-path vulnerability affecting versions through 2026.1.
  • N-able releases Hotfix 1 (build 2026.3.1.7) and publishes its first security advisory; analysis confirms an alternate exploitation vector (CVE-2026-18577) bypassing the earlier fix. Finland's national cyber security centre also issues an advisory.
  • A later same-day Huntress tracking update shows improving patch compliance: 13.6% of internet-reachable cloud N-central servers and 28.6% of self-hosted instances still unpatched/vulnerable.
  • Huntress publishes a rapid-response writeup confirming exploitation across multiple organizations -- in one case a single compromised self-hosted MSP partner account reached nine downstream client organizations, one endpoint compromised per organization -- and finds 55.6% of internet-reachable cloud N-central servers still unpatched as of 12:45 AM ET.
  • CISA adds CVE-2026-18577 to its Known Exploited Vulnerabilities catalog with an August 6, 2026 remediation due date under BOD 22-01.
  • Sophos identifies a compromised customer in its telemetry at approximately 08:00 UTC, later attributing the activity to CVE-2026-18577 exploitation.
  • CISA adds CVE-2026-18556 to its Known Exploited Vulnerabilities catalog (due August 7, 2026) after N-able's follow-up advisory.
  • Sophos independently identifies a second compromised organization in its telemetry.
  • N-able publishes a custom N-central service template that automates IOC detection on managed Windows endpoints, cautioning a clean scan does not guarantee an environment is unaffected.
  • N-able releases Hotfix 2 (build 2026.3.1.10), stating it is required even for customers who already applied Hotfix 1, and discloses an expanded 10-IP indicator list.
  • Follow-up reporting confirms attackers reached managed downstream systems via the Take Control feature, created a rogue domain account, deployed a BYOVD-style EDR-evasion tool (locked-ioc-20/locked-ioc-35) to disable Sophos protection, and persisted via a masqueraded Cloudflare Tunnel service even after N-central server access was revoked.

Sources cited for N-central Cloudflare Tunnel persistence

Showing 15 of 22 sources, primary sources first.

Drafted by the Threadlinqs research pipeline from 22 sources; published ; last updated —.

Threadlinqs Intelligence — Real-Time Threat Detection Platform

[ 0 threats ] [ 0 det ] [ CRIT: 0 ] [ HIGH: 0 ]
// threat_feed
$ sort --newest
Showing all threats

Live intelligence console

Threat level

Latest Threats